June 2004: Part 4

June 2004: Part 4

Previous Top Next

8/6/2004
Oxide layers
Using DD1 we produce three types of oxide. With metallic targets and the highest possible oxygen flow, we try to produce stochiometric A-oxide, B-oxide and C-oxide layers. According to literature, these oxides should have different refractive index values that we will need for our coating products. In this first test, for each oxide type we produce a thin and a thick layer (on microscope slides).
Transmittance and reflectance spectra are recorded with the S1 spectrometer and stored to the database. Before we record the oxide layer spectra, we take spectra of a bare substrate in order to check if everything is Ok with the measurements. The substrate spectra are stored in the database directory of A-oxide.

13/6/2004
Optical constants of the oxides
Today we try to determine the optical constants of the produced oxides. A visual inspection of the measured spectra shows that in the case of A-oxide the spectra are not very different from the substrate spectra. This looks a little difficult, so we decide to start with B-oxide. Clearly the spectra of the thick layer exhibit more structures (we think we see an interference pattern) and we decide to start the analysis with the thick layer.
M.Theiss Hard- and Software (the supplier of the CODE software) sent us a CODE configuration that they recommend for oxide analysis. They customized a general solution based on the so-called 'Universal oxide' model. The configuration looks like this:

The optical constant model consists of three susceptibility contributions:

1.	A constant (called 'Dielectric background') describes the response of electronic excitations in the x-ray and deep UV region, far away from the spectral range of interest in this case. This contribution does not lead to any absorption or dispersion.

Interband transitions in the far UV are represented by a single harmonic oscillator model. The model has three parameters: the resonance frequency, the oscillator strength and the damping constant (see SCOUT technical manual). We use this oscillator to 'summarize' interband transitions outside the investigated spectral range - it need not describe any details of these transitions. The resonance frequency is fixed at 60000 1/cm (167 nm wavelength). The low and fixed damping of the oscillator (100 1/cm) ensures that no absorption is caused by this term in visible spectral range. However, this contribution causes dispersion in the spectral range we are interested in.

For the interband transition with the lowest energy we use the OJL model which has turned out to be excellent for amorphous materials. This model will be able to introduce both tail state absorption and absorption by electronic transitions from the valence to the conduction band. (both assumed to be parabolic). The corresponding dispersion of the real part of the complex refractive index is included as well.

Using the sliders in the CODE configuration we play with different strengths of the individual contributions to the optical constants and inspect the consequences for the optical properties of the material. First we switch off the 2nd and 3rd term and work with a constant refractive index only. With a proper layer thickness we find good agreement of model and measurements in the infrared:

However, above 500 nm a higher refractive index is needed, keeping the infrared level of the refractive index the same. This dispersion can be achieved by increasing the 'UV oscillator strength' and decreasing the 'Dielectic background':

To inspect the remaining problems of the model in the UV reflectance we switch to a logarithmic scaling of the wavelength axis for reflectance:

In order to improve the fit in this range the OJL model has to be applied. Increasing its strength (while decreasing the dielectric background and the oscillator strength) and adjusting manually the bandgap a rather good approach is found:

The model is now ready for an automatic parameter fit. We select the layer thickness, the dielectric background, the UV oscillator strength and the strength, the bandgap and the tail state exponent of the OJL model as fit parameters. Important: The tail state exponent must not be smaller than 0.01 (otherwise you risc a nasty crash of CODE) - so we set a lower and an upper limit for the tail state exponent in the list of fit parameters.
In addition, it turns out to be useful to increase the number of data points (use the global Range command in the main window) from 200 to 500 in order describe the sharp interference structures in the reflectance between 200 and 300 nm. The final fit looks rather good:

This configuration is saved in the database folder 'software configurations/code' using the filename B_oxide_fit_thick_layer.wcd.

Before we save the optical constants for B-oxide to the database we check if the thin film sample can be described as well with the current model. Using the Import command in the main window we load the thin film spectra of B-oxide. After freezing all optical constant parameters in the fit we select the grid fit option for the layer thickness (see SCOUT technical manual, boundaries for the thickness: 0 ... 500 nm, grid with 100 points). Immediately CODE finds good agreement with a thickness of 25.1 nm:

clip0024

This is convincing! Both the thin and thick layer of B-oxide can be described with the same optical constant model. We save the model to the database with an appropriate comment. Let's hope that we can have similar success for the A- and C-oxides as well.

14/6/2004
After yesterday's success, today we hope to get good results for the other oxides as well.

A-oxide
In the case of A-oxide, we already noticed that the spectra are similar to the substrate spectra. Therefore, very likely A-oxide has optical constants similar to those of glass. After some manual slider adjustments the fit goes well for the thick layer:

The corresponding thin film fit without further modifications of the optical constant model is satisfying but there is a small disagreement in the reflectance:

clip0028

A slightly higher refractive index would improve the fit a little. The easiest way to achieve this is to 'unfreeze' the dielectric background and fit it together with the thickness. The result is the following:

clip0029

The slightly increased refractive index could mean that the thin layer is - on the average - a little more compact than the thick layer. In any case, we store both optical constant versions of A-oxide to the database.

C-oxide
The fit for the thick C-oxide layer is easy after some manual adjustments (reducing the bandgap, for example):

The agreement of the thin film spectra is satisfying. No further modification is required in this case: